FPS and 1% lows: looking beyond the average

A game creates the impression of movement by showing images in succession. FPS stands for “frames per second”. At 60 FPS, the game produces an average of 60 frames each second; at 120 FPS, it produces twice as many.

La moyenne est un bon premier repère, mais elle masque les variations. Un jeu peut afficher un chiffre élevé et donner malgré tout l’impression d’accrocher. Comme une vitesse moyenne en voiture : elle ne raconte pas tous les coups de frein du trajet.

The “1% low” focuses on the slowest part of the measured sequence. Expressed in FPS, it helps identify drops that the average can hide. Under identical conditions and with the same calculation method, a higher value is generally preferable.

It is neither the absolute minimum nor a promise that the game will never drop any lower. Depending on the software, this number can represent a threshold separating the slowest frames from the rest, or an average calculated over a slow portion of the capture. You therefore need to compare results calculated in the same way.

Let us use a fictional example, solely to explain how to read the results:

• Configuration A: 100 FPS on average, with a 1% low of 75 FPS.

• Configuration B: 100 FPS on average, with a 1% low of 40 FPS.

The average puts the two configurations level. Yet with the same scene, settings and calculation method, configuration B shows more pronounced slowdowns in the slow part of the measurement. Configuration A performs better on this metric here.

These two numbers do not tell you when the slowdowns occur, however, or whether they cluster together. To understand that, you can look at frame times.

“Frame time” is the measured time per frame, in milliseconds. At a perfectly steady rate, 60 FPS corresponds to around 16.7 ms per frame; 120 FPS corresponds to around 8.3 ms.

On a frame-time graph, an upward spike indicates a frame taking longer than the preceding ones. If the timing suddenly jumps from 16.7 to 80 ms, that wait can feel like a hitch. An average calculated over several minutes can almost hide it entirely.

When reading this graph, look for short, consistent times. A flat line alone is not enough: at 50 ms per frame, the pace is steady, but it amounts to just 20 FPS.

Before choosing between two graphics cards, check the resolution: Full HD, 1440p or 4K. More pixels generally mean more work. Check the graphics settings too: “medium” and “ultra” do not demand the same effort.

The processor, game version, graphics driver and tested sequence also matter. Walking down an empty corridor and crossing a crowded city are not the same test. To compare with your own PC, reproduce the conditions stated in the article as closely as possible.

Watch out for frame generation too: it adds intermediate frames to those rendered by the game. It can improve visual smoothness, but doubling the FPS counter does not automatically mean your controls respond twice as fast. Compare results with and without this option separately, and also check image reconstruction modes such as DLSS or FSR.

Start with the game, resolution and settings you are interested in. Then look at the average, followed by the 1% low when provided. If two averages are close, this second metric can reveal a difference that the first bar does not show.

For FPS values, higher is generally better. For times in milliseconds, lower is generally better. The review commentary then explains any observed stuttering and the limits of the comparison.